Research And Implementation Of Gear Profile Measurement Method Based On Machine Vision

Gear is a kind of mechanical transmission component widely used in varieties mechanical transmission system.The manufacture ability level of a country can be partly reflected by how they produce accurate gear.The scale of China's gear market and the increasing requirements for high precisious gear in various applications promote an upgrade in gear production technology as soon as possible.Gear measurement is an important part of the gear production process.However,the current gear measurement equipment is high precision but low efficiency and high cost.These features leave a big problem to both small companny in cost and operator to use.Therefore,in this paper I proposes a gear measurement system based on embedded equipment with machine vision.This measurement system has the following characteristics: 1.Low cost.Compared with the tens of thousands of image measuring instruments and hundreds of thousands of three coordinate measuring machines(CMM),cost of the measuring system is just three thousands RMB.2.Easy to use.Human-computer interaction interface is similar to a mobile app easy to use for operators.3.Moderate accuracy.The measurement adopts the visual measurement method.Although the measurement accuracy is not as good as the CMM,it can meet the measurement of gear with the accuracy 7 ~12 level widely used in daily life.The main content in this paper is surrounding the above three characteristics to introduce the design and implement of this embedded vision measurement system:Firstly,Aftrer comparing the current several embedded hardware platforms architecture,The Raspberry Pi 3B based on Broadcom BCM2837 chip is selected as the hardware development platform.Because we can use hardware,VideoCore GPU intergraded in BCM2837,to accelerating the graph algorithm in our measure to achieve a balance between performance and price.To reduce the error caused by the assembly of our camera,a simple but useful analysis method,analysising the pixel coordinates of the center of the calibration plate at different object distances,be proposed.Then,Base on the open source platform Qt and self-developed Raspberry Pi 3B GPIO interface driver,i imple a simple and practical graphical user interaction system.For the image algorithm,the common image graying algorithms are firstly disscussed.According to the application characteristics of the measurement system,an improved RTCP algorithm is proposed.After studied several sub-pixel measurement methods such as Gaussian fitting method,curve interpolation method,moment method,finally I decided to combine with Canny pixel edge detection algorithm and Zernike moment algorithm to achieve sub-pixel edge detection to ensure measurement accuracy.Considering the defect of least squares fitting the center of a circle,this paper uses the method of minimum radial deviation sum to correct it,and then applies the algorithm to the measurement system based on OpenCV.In the end,by the measurement experiment,the feasibility of the measurement system is verified.